What follows are excerpts from Lee and Webster’s wide-ranging responses to the CISTO News (CN) interviewers’ questions.
CN: How is the HEC Program viewed among your Headquarters colleagues?
Lee: From the Earth science side, it is paramount. We cannot do that kind of research without a computational capability. Moreover, we are seeing increasing utilization from the space science side of the Science mission. In addition, our Aeronautics mission is focusing more on fundamental aeronautics, and much of that work is computationally intensive. Similarly, when they design a vehicle, Exploration needs to use very complex models to simulate vehicles through different flow regimes. So, HEC is very, very important to all NASA missions.
CN: Are there any missions that need high-end computing in space? The James Webb Space Telescope, which wants to unfold a large aperture and align it properly, comes to mind.
Lee: Right now, there has been a lot of discussion of high-end computing in space. That is related to data assimilation, trying to extract information from the observations and send the minimum amount of bits down to the Earth. Since you mentioned unfolding an aperture, in fact we are looking into some engineering problems. For example, when SOFIA’s [Stratospheric Observatory for Infrared Astronomy] aperture creates a cavity, what kind of aerodynamic impact is there on the spacecraft? Also, there are interesting ideas to use SOFIA to observe near-Earth objects. We need to block or shield the starlight, and there has been discussion of using supercomputing to simulate that kind of shield.
CN: How robust is the Headquarters funding climate with respect to HEC?
Lee: Actually, it is quite stable. We had an increase in funding last year, and we are trying to maintain that. With Headquarters funding, NASA programs, overall, are under a lot of stress. High-End Computing is one area that we are trying to protect. That is a very significant commitment to maintaining a robust HEC capability.
CN: What recent or current HEC applications have had a major impact on NASA missions? In the near-term, what high-impact applications do you envision?
Webster: I see a tremendous amount of work that has come out of Jim Hansen’s Goddard Institute for Space Studies and their IPCC [Intergovernmental Panel on Climate Change] simulations. A lot of interesting and ground-breaking work has come out of Joan Centrella’s group with the black hole research they have been doing. Climate model development in the GMAO [Global Modeling and Assimilation Office] is very interesting and will have a large impact.
Lee: Exploration and Space Operations do a lot of engineering and CFD [computational fluid dynamics] research. Also, they are looking at how to design heat shields and use different materials that will sustain flight through the very difficult environment of space. These materials go through very different flow regimes. Furthermore, when you put people in the Crew Exploration Vehicle, how do you protect them from radiation? Many computing cycles are necessary to simulate space radiation and its interaction with spacecraft materials.
CN: What do you anticipate or hope will be the major scientific or engineering advances that result from NASA HEC?
Webster: Thirty or forty years ago, people had two tools to do research and engineering. One was theory, and the other was experimental observations. And now, with the advent of HEC, we have the ability to do simulations with incredibly high temporal and spatial resolution. I believe that we will be able to fully simulate what is happening with the climate. The exact causes of climate change and what, if any, mitigating actions should be taken are different than NASA’s scientific mission of being able to predict. That becomes a political or a policy issue for other people within the federal government. NASA is an incredibly great place to work, and it has accomplished incredible things in the past. NASA has a very ambitious agenda of exploration and research. I believe that high-end computing is going to be a critical component of the entire Exploration mission. I don’t see how we can develop vehicles, missions, and robots to go to the far corners of the universe without simulation on high-end computers.
Lee: With more and more observations at higher resolution, I see an opportunity coming up when we use high-resolution simulations and data assimilation to validate or to prove theories. I see that new, significant use of high-end computing in the future.
Webster: There are many fundamental questions of basic physics that, at some level, have to be modeled. If you look at Joan Centrella’s research, the basic interactions of monstrous black holes, there are fundamental things in there that can only be fully understood with using high-end computing.
Lee: With Joan Centrella’s work, a major question being addressed is if we can simulate the observations and try to predict what we can observe once we put LISA [the Laser Interferometer Space Antenna] in space. Similar work is starting to happen in Earth science with OSSE, the Observation System Simulation Experiment. You simulate your observation in a virtual world and use that as a design and validation tool.
CN: What is your vision for a future NCCS and HEC Program?
Webster: My vision for the future of the NCCS is what I call a data-centric computing center. High-performance computing centers in the past have been primarily processor-centric. Within the NASA community, new knowledge comes from the data. So, I’m trying to build the NCCS around data and data services that allow researchers to interact with and interrogate very large data sets, whether generated numerically or experimentally derived.
Lee: Data are really what we are about at NASA. We put up a lot of space assets, we collect data, and we generate a lot of data, so we have to focus on data. In addition, if we look at history, supercomputing centers did a lot of research building new supercomputers. Now, high-end computing is very quickly becoming a commodity. In the future, we need to develop expertise in acquisition, service, management, and exploring new offerings from the market instead of building our own systems.
CN: As PC clusters become more powerful and affordable, how will this impact traditional HEC as viewed by NASA? That is, do you anticipate a continuing need for centralized nodes at Ames and Goddard?
Webster: Historically, a tremendous amount of work was done on workstations, but it was always limited to the size of the workstation. Now, I believe these PC clusters have nominally replaced the workstation, but they still will not approach the size of problems that will be supported on supercomputers. So, supercomputers or high-end computers will always be there and will always be beyond what a single cluster, user, or project can do. Our scientists will demand a computer at that level for the advancement of science.
So, what is the impact on NASA? These PC clusters are still being built out of commodity processors, but they are growing. They are incredibly power-hungry, and they generate a lot of heat, which means that you can’t just put them anywhere anymore because they put a tremendous burden on a building’s infrastructure. So, if nothing else, a place like the NCCS or NAS at Ames has the infrastructure to support these types of computers. As a matter of fact, our limiting issue is the amount of cooling available for supercomputers that are projected to be deployed in the country in the coming years. Bottom line, we are driven to the point that we must use the components the industry provides for us. However, we will always be configuring these machines well beyond the design envelope that they were designed for, and hence the challenge of supercomputing.
Lee: I don’t really see PC clusters, per se, becoming widely used. But the type of systems Phil just described, you take the most capable server from the market and try to configure it beyond what the commodity market will do. You take 16- or 32-way servers and put them together and make a bigger cluster. That is the way we push the limit. The time has passed for us to design our own computer from the ground up. We cannot afford to do that anymore.
Also, if you look at technology advancement, it is following Moore’s Law [the number of transistors in an integrated circuit doubles every 2 years, with component cost staying the same] and exponentially increasing. If you think of human intelligence, we are not advancing or growing at that kind of speed. It is challenging for scientists and engineers to design applications to take full advantage of the most sophisticated systems. For NASA, that challenge hinges on whether or not you can analyze the data, make sense of the data, and also find the observations to match up with what you can simulate. A computer is a very good tool to generate data by itself, but you have to constrain that data with your observations. If you have the most sophisticated simulation but don’t have data to validate or constrain it, what is good about it?
CN: What value beyond computing cycles does HEC provide to its users?
Webster: If you look at how these computers are being built, we are at a paradigm shift, moving from ever-increasingly powerful single processors to multiple processors all running at the speed of the old ones. There are a lot of issues about how users program these multi-core processors and, as Tsengdar described, how they deal with and evaluate the data. So, HEC needs to be able to supply knowledge to users about how to program these machines. We also have to provide a larger analysis environment and a visualization environment, things we have not traditionally done. As the problems get bigger and the data sets get so large, we have to present different tools to the user to interrogate these data.
CN: Some say that the network is the computer. How important is this issue to NASA HEC?
Lee: It is absolutely critical. If we look at the direction of HEC in the use of commodity systems and the building up of the clusters, the network is used for interconnect inside the cluster. The network also connects the compute systems to analysis nodes and mass storage available locally—as well as remotely. The network will be especially critical in the future, when not only will the compute nodes be distributed but also the data. You will have data sets in different locations, and you will have to bring those data to the analysis nodes.
Webster: I agree philosophically with the statement, but I don’t believe the statement is correct that the network is the computer. Factually, you would have to say that the network is the backplane of the computer. If you compare your standard workstation, which has disk, screen, mouse, and processor, it also has to have a backplane that connects them. So, all the things Tsengdar just talked about, with a high-speed network as the backplane, now you can have the components of the computer geographically dispersed.
CN: What benefits to HEC users do you foresee from the transition to the NASA Integrated Services Network (NISN) away from the current NASA Research and Engineering Network (NREN)?
Lee: It is an economy-of-scale kind of argument. The Agency has a service-provider in NISN, so why don’t we leverage it instead of developing our own dedicated resource? In terms of the service aspects, NREN and NISN are not too different in their offerings. Of course, there are different technologies, and we can argue whether this technology is better than that one. But, strategically, NASA is in a more budget-constrained environment. We have to leverage other organizations’ offerings, just as we share our own resources with other organizations.
CN: Are you satisfied with the organizational structure of HEC at both Ames and Goddard? Do you anticipate the need for any reorganization?
Lee: That is really up to the Centers.
Webster: Certainly within Goddard, I believe that the NCCS will probably need to be more aligned in the future with the Applied Engineering and Technology Directorate, Code 500. We have a very robust engineering activity here. I suspect that there will be reorganizations of our computing center, both for alignment with the missions as well as the changing focus of the Information Technology and Communications Directorate, Code 700.
CN: Is the Engineering Directorate knocking on your door for compute services?
Webster: Not yet, but they have only recently started knocking on Tsengdar’s door as Exploration has fulfilled itself. They now have a strategic plan, and they are starting to develop a research agenda and programs. I expect that, as Goddard becomes more competitive for a lot of these future missions, high-end computing will become necessary. If I just look at the pioneering work that Rick Lyon is doing in computational optics, the next step is modeling the different parts of a spacecraft.
http://www.hec.nasa.gov
|
